Workshop Diary is an ongoing series about how tools actually get built, in my own projects and my clients’.
I take backups of my site, like of everything else. And for years I made the same mistake almost everyone makes: I thought taking the copy was the job. It isn’t. The job is getting your stuff back, and those are two completely different things. A backup you’ve never restored anything from is not a backup. It’s a hope. This is a diary entry about that difference, and about how you usually learn it: namely, the hard way.
The difference few people think about
Backing up is really two separate things treated as if they were one. The first is making the copy: the data is copied somewhere safe. The second is the restore: getting the data back into use when the original is gone. Almost all the attention goes to the first, is the copy running, is it scheduled, is the light green. And almost no one checks the other half: does the restore actually work.
This is backwards, because the restore is the only part that matters. The copy is just a means. The restore is the end. Nobody actually wants a backup, people want their stuff safe and available. A copy without a working restore doesn’t give you that. It only gives you the feeling of it, like an insurance policy no one has read.
Why an untested copy fails exactly when you need it
Here are the ways an untested backup falls apart, and they all reveal themselves at the worst possible time. The copy may have been saving the wrong folder, so what you thought was there isn’t. The files may be corrupt, and you don’t notice until you try to open them. The copying may have quietly stopped months ago, and no one noticed because nothing shouted. The copy may be safe, but you don’t have the password or key to open it. The restore may take three days you don’t have. Or the copy has faithfully saved the same problem too, the encrypted file, the broken database, the malware. That put you in trouble in the first place.
What these have in common is that they only come to light when you try to restore. And if you try to restore for the first time only when a real crisis is on, you’ve chosen the worst possible moment to learn that your copy didn’t work.
The moment you find out
I know that moment, and probably you do too. The disk dies, the site breaks, or the message appears on screen that your files are encrypted. The first feeling isn’t panic but relief: ”good thing I took backups”. You bring out the copy, calm. And then the restore doesn’t work, the file is empty, the key is missing, or half the data wasn’t included. That calm relief turns to cold dread in a second. That is exactly the moment an untested backup reveals it was never a backup. And by then there’s nothing left to do about it.
How to do it right: test the restore
The solution is embarrassingly simple, and yet almost nobody does it: actually restore from your backup. Take it to a separate place, open the files and check that they work. Not once, but regularly. That’s the whole thing. It’s no more complicated than that. It just requires that you stop assuming and start checking.
This is exactly the same lesson as with practice: a plan you haven’t rehearsed is just a wish in written form. A backup is a plan for disaster. An untested backup is an unrehearsed plan. It might work, but you don’t know that, and ”might work” is the worst sentence in all of preparedness, because it’s always said right before something doesn’t.
What’s actually worth checking
When you test the restore, look at these five things. First, that you get the files restored to a separate place and they actually open. It’s not enough that they appear in a list. Second, that the copy is saving the right things: a surprising number of backups diligently copy an empty or wrong folder. Third, that you have everything you need for the restore, passwords, keys, the right software, stored somewhere other than the broken system itself. Fourth, that the restore is fast enough: if it takes three days and you need the data tomorrow, the copy isn’t enough. And fifth, that you can go back in time past the point of the problem, meaning you have copies from several points, not just yesterday’s, which may already contain the same fault.
An old rule of thumb sums up the basic structure: three copies, on two different media, one in another place. That’s a good frame. But even it is worthless without testing, three untested copies are just three hopes instead of one.
Why this applies to absolutely everything
Even if you never back up anything, here’s the same law that recurs throughout preparedness. A backup is continuity management in miniature: it’s a plan for when something breaks. And every plan is subject to the same merciless rule, the system, the copy, the contingency plan you’ve never tested will fail exactly when it’s supposed to save you. A company’s continuity plan, a standby generator, an alert chain, a home emergency supply, same thing. Existence isn’t the same as working, and the only way to know the difference is to try it before you have to.
Finally
I test my restores now. It feels like a waste of time when nothing is broken, exactly the way a fire drill feels pointless when nothing is burning, or a search exercise pointless when no one is lost. But that ”wasted” hour is the whole point. It’s the difference that separates the sentence ”I have backups”, which is a hope, from the sentence ”I can get my stuff back”, which is a fact. And you notice that difference in only two ways: either by testing calmly in advance, or the hard way in the middle of a crisis. The latter is always more expensive. So test yours. Today, not when the disk has already died.